CN104569907B - Wireless positioning method and system based on neural network and road side unit - Google Patents
Wireless positioning method and system based on neural network and road side unit Download PDFInfo
- Publication number
- CN104569907B CN104569907B CN201410447334.3A CN201410447334A CN104569907B CN 104569907 B CN104569907 B CN 104569907B CN 201410447334 A CN201410447334 A CN 201410447334A CN 104569907 B CN104569907 B CN 104569907B
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- gathered
- output
- neuron
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received
- G01S3/02—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using radio waves
- G01S3/74—Multi-channel systems specially adapted for direction-finding, i.e. having a single antenna system capable of giving simultaneous indications of the directions of different signals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a wireless positioning method and system based on a neural network and a road side unit. The wireless positioning method comprises the following steps: acquiring wireless signal data from a target, wherein the acquired data comprises amplitude, phase and/or power of a wireless signal; preprocessing the acquired data; inputting the preprocessed data into a pre-established neural network model and taking output of the neural network model as a positioning result of the target. By implementing the technical scheme of the invention, a tiny difference of signals can be adaptively identified and a relatively accurate positioning result can be given.
Description
Technical field
The present invention relates to wireless positioning field, more particularly to it is a kind of based on the wireless location method of neutral net, system and
Roadside unit.
Background technology
The orientation of estimation space signal is processed to the phase place or amplitude of wireless signal currently with aerial array, from
And the method for obtaining the position of wireless signal correspondence target mainly has 1.MUSIC (Multiple Signal
Classification Multiple Signal Classifications) algorithm and innovatory algorithm, 2. phase method coherent signal ambiguity solution and innovatory algorithm,
3. the statistical method of probability is based on.Above-mentioned many algorithms are all based on a certain class criterion, go to estimate the wireless of maximum possible appearance
Signal angle direction, although respectively have pluses and minuses.But go to estimate due to itself being all based on a certain criterion, so for small
Difference is all helpless.
The content of the invention
The technical problem to be solved in the present invention is, for prior art above-mentioned positioning when None- identified minute differences
Defect, there is provided a kind of based on the wireless location method of neutral net, system and roadside unit, can recognize small difference in positioning
Not.
The technical solution adopted for the present invention to solve the technical problems is：A kind of wireless location based on neutral net of construction
Method, including：
S1. the wireless signal data from target are gathered, the data for being gathered include the amplitude and/or phase place of wireless signal
And/or power；
S2. the data to being gathered carry out pretreatment；
S3. by pretreated data input to the neural network model for pre-building, and by the neutral net mould
Positioning result of the output of type as target.
In the wireless location method based on neutral net of the present invention, neural network model is built according to the following steps
It is vertical：
S31. the number of input neuron, the number of the data to be gathered are determined according to the number of the data to be gathered
It is related to the number of bay；
S32. the number of output neuron is determined；
S33. the number of plies and intermediate layer neuron number in the middle of neuron is determined；
S34. the wireless signal data from known target are gathered in training, and the data to being gathered carry out pre- place
Reason；
S35. the pretreated data of a part of step S34 institute are used for training neural network model, and by constantly repairing
Change each interneuronal weight, until the neural network model meets the convergence of preset rules.
In the wireless location method based on neutral net of the present invention,
When the output type of the neural network model is positioning region, by the region division that need to be positioned into M preset areas
Domain, single output neuron has two class output states, therefore the output of neural network model has 2NClass output state, wherein N
For the number of output neuron, the output state per Connectionist model is empty or corresponding with a predeterminable area, if defeated
Do well corresponding with a predeterminable area, that is, determine target in corresponding predeterminable area.
In the wireless location method based on neutral net of the present invention,
The output type of the neural network model for relative coordinates when, the number of the output neuron is 2, described
Output neuron is output as coordinate figure.
In the wireless location method based on neutral net of the present invention, step S2 includes：
Wireless signal to being gathered carries out pretreatment, obtains instant amplitude value；Or；
The mean power for carrying out pretreatment, obtaining the wireless signal of the wireless signal to being gathered in Preset Time
Value；Or；
Wireless signal to being gathered in Preset Time carries out pretreatment, obtains wireless signal between different antenna element
Phase contrast.
The present invention also constructs a kind of roadside unit, including：
Acquisition module, for gathering the wireless signal data from electronic tag, the data for being gathered include wireless signal
Amplitude and/or phase place and/or power；
Pretreatment module, for carrying out pretreatment to the data for being gathered；
Processing with Neural Network module, for by pretreated data input to the neural network model for pre-building,
And using the neural network model output as carrying electronic tag vehicle positioning result.
In roadside unit of the present invention, the Processing with Neural Network module includes：
Input module, for determining the number of input neuron according to the number of the data to be gathered, to be gathered
The number of data is related to the number of bay；
Output module, for determining the number of output neuron；
Intermediate module, for determining the number of plies and intermediate layer neuron number in the middle of neuron；Adjusting module, for using
The wireless signal data training neural network model of collection and the pretreated electronic tag from known location during training, and
By constantly changing each interneuronal weight, until the neural network model meets the convergence of preset rules.
In roadside unit of the present invention, the acquisition module includes：
Aerial array, for receiving the wireless signal from electronic tag, the aerial array is phased array antenna；
Radio-frequency transmitter, for being demodulated to the wireless signal that the aerial array is received；
Analogue signal processor, for processing the signal after radio-frequency transmitter demodulation；
Analog-digital converter, for carrying out analog digital conversion to the signal after processor process.
The present invention also constructs a kind of wireless location system based on neutral net, including：
Acquisition module, for gathering the wireless signal data from target, the data for being gathered include the width of wireless signal
Degree and/or phase place and/or power, the acquisition module is multiple antennas；
Control module, for the data for being gathered to be carried out with pretreatment, and by pretreated data input to building in advance
In vertical neural network model, and using the output of the neural network model as target positioning result.
In the wireless location system based on neutral net of the present invention, the control module includes：
First input module, for determining the number of input neuron according to the number of the data to be gathered, to be adopted
The number of the data of collection is related to the number of bay；
First output module, for determining the number of output neuron；
First intermediate module, for determining the number of plies and intermediate layer neuron number in the middle of neuron；
First adjusting module, for using collection and the pretreated wireless communication count from known target in training
According to training neural network model, and by constantly changing each interneuronal weight, until the neural network model meets pre-
If the convergence of rule.
Implement technical scheme, by using neutral net (artificial neural network, ANN)
Technology, the wireless signal data to being collected carry out fusion treatment, and a shade of difference of self-adapting estimation signal provides more accurate
Positioning result.
Description of the drawings
Below in conjunction with drawings and Examples, the invention will be further described, in accompanying drawing：
Fig. 1 is flow chart of the present invention based on the wireless location method embodiment one of neutral net；
Fig. 2 is the flow chart of Establishment of Neural Model embodiment of the method one；
Fig. 3 A-3E are respectively the relation schematic diagrams of different output neuron numbers and object location result；
Fig. 4 is the schematic diagram of three array-element antenna arrays；
Fig. 5 is the schematic diagram of weight embodiment one between each neuron in neural network model；
Fig. 6 is the logic chart of roadside unit embodiment one of the present invention；
Fig. 7 is a kind of logic chart of the wireless location system embodiment one based on neutral net of the present invention.
Specific embodiment
Chinese patent ZL200480025276.X discloses a kind of method for building up of artificial neural network, artificial neural network
(artificial neural network, abridge ANN), abbreviation neutral net (neural network, abridge NN), is near
A kind of mathematical model or computation model of the 26S Proteasome Structure and Function of the mimic biology neutral net risen over year.Neutral net is by a large amount of
Artificial neuron be coupled calculated.In most cases artificial neural network can change internal on the basis of external information
Structure, is a kind of Adaptable System.Modern neuro network is a kind of Nonlinear Statistical data modeling tool, is commonly used to input
Complicated relation is modeled with outlet chamber, or for the pattern of heuristic data.
By the learning capacity with artificial neural network, can cause to position target location self adaptation by aerial array
The small difference of identification, so as to improve accuracy of identification.
Fig. 1 is flow chart of the present invention based on the wireless location method embodiment one of neutral net, should be based on neutral net
Wireless location method include：
S1. the wireless signal data from target are gathered, the data for being gathered include the amplitude and/or phase place of wireless signal
And/or power, i.e. the data for being gathered can include one, or amplitude, phase place, work(in amplitude, phase place, power
The combination in any of rate；
S2. the data to being gathered carry out pretreatment, and the pretreatment of data for example includes data scrubbing, data integration, number
According to conversion, data regularization etc., in one embodiment, the amplitude of wireless signal that can be to being gathered carries out pretreatment, to obtain
Instant amplitude value；Or, the phase place of the wireless signal to being gathered in Preset Time carries out pretreatment, to obtain different antennae battle array
The phase contrast of wireless signal between unit, and pretreatment is carried out to the power of wireless signal that gathered in Preset Time, obtain
The average power content of the wireless signal.In addition, many array-element antenna arrays gather respectively the wireless signal sent from target
Phase data, in pretreatment, the difference of the phase data that any two array element is gathered is calculated respectively, then can also divide
The other difference to these phase datas is normalized；
S3. by pretreated data input to the neural network model for pre-building, and by the neutral net mould
Positioning result of the output of type as target.
Fig. 2 is the flow chart of Establishment of Neural Model embodiment of the method one, the neural network model of the embodiment according to
The following steps are set up：
S31. the number of input neuron, the number of the data to be gathered are determined according to the number of the data to be gathered
It is related to the number of bay；
S32. the number of output neuron is determined；
S33. the number of plies and intermediate layer neuron number in the middle of neuron is determined；
S34. the wireless signal data from known target are gathered in training, and the data to being gathered carry out pre- place
Reason；S35. the pretreated data of a part of step S34 institute are used for training neural network model, and by constantly changing each god
Weight between Jing is first, until the neural network model meets the convergence of preset rules.
In step S31, it is determined that the number of data that the number of input neuron can be gathered as needed, for example, three gusts
The phase data of first aerial array collection has three, and three for being gathered phase data can be converted into after pretreatment three phases
The data of potential difference, to gather two data.The number that can determine that input neuron is two, by this any two phase contrast
Data input is to this two input neurons.From the foregoing, it will be observed that when array element is 3, and the data of collection are phase contrast, input nerve
Up to 3, unit, naturally it is also possible to less than 3.When the data of collection are phase place or power, or three kinds when arbitrarily combining, its
The relation of input neuron and array element also complies with above-mentioned rule.
In step s 32, determine that the number of output neuron can be according to the actual demand of user.For example, fixed to vehicle
During position, the OBU (On Board Unit, board units) on vehicle to RSU (Road Side Unit, roadside unit) sends nothing
Line signal, RSU carries out pretreatment by the wireless signal to being received and artificial neural network is processed, and can orient the position of vehicle
Put.In this process, the output type of neural network model can be positioning region, alternatively relative coordinates.
It is single by the region division that need to be positioned into M predeterminable area when the output type of neural network model is positioning region
Individual output neuron has two class output states, therefore the output of neural network model has 2NClass output state, wherein N are output
The number of neuron, the output state per Connectionist model is empty or corresponding with a predeterminable area, if output state
It is corresponding with a predeterminable area, that is, determine target in corresponding predeterminable area.
For example, in one embodiment, as shown in Figure 3A, 1. still exist in this track region if need to only orient vehicle
Adjacent track region 2., now, it may be determined that go out output neuron number be one.When the output neuron output 0, car is represented
In this track region 1., when output neuron output 1, represent vehicle in adjacent track region 2..In another embodiment
In, as shown in Figure 3 B, specific region 1. (transaction location of such as setting), this car of vehicle in this track is oriented if desired
Other two regions in road 2., 3. (for example sail into region and roll region away from) or adjacent track region 4., now, it may be determined that go out output
The number of neuron is two, when the two output neuron output 00, represents vehicle in this track region 1.；When this two
During output neuron output 01, represent vehicle in this track region 2.；When the two output neuron output 10, vehicle is represented
In this track region 3., when the two output neuron output 11, represent vehicle in adjacent track region 4..Similarly, at it
In its embodiment, as shown in Figure 3 C, which in eight regions classified of vehicle is oriented if desired, it may be determined that go out
The number of output neuron is three.As shown in Figure 3 D, vehicle is oriented if desired in 20 regions classified
Which, it may be determined that the number for going out output neuron is five, it should be noted that, it is five in the number of output neuron, institute
When the region of division is 20, a part of value correspondence inactive area that five output neurons are exported or multiple values correspondence are together
One region.
The number of positioning region, the number of output neuron, the corresponding relation such as institute of table 1 of the output valve of output neuron
Show：
Table 1
When the output type of neural network model is relative coordinates, the number of output neuron is two, the output
Neuron is output as coordinate figure.Relative coordinates set up mode：RSU hangs down point for origin, and track is Y-axis, crosses origin and and car
Road it is vertical for X-axis.In two output neurons, an output neuron is used for exporting x coordinate value, another output nerve
Unit is used for exporting y-coordinate value, as shown in FIGURE 3 E, the positioning result of the output of two output neurons for (1.2,4.6).
In step S33, the number of plies and intermediate layer neuron number in the middle of neuron are determined, that is, determine intermediate layer neuron.
The number of plies is generally less than equal to 3 in the middle of neuron, and in theory, the middle number of plies is more much more accurate.Can according to circumstances carry out in practice
Adjustment.Table 2 is a reference table of the neuron intermediate layer number of plies and intermediate layer neuron number：
Table 2
In step s 35, in step S34 the collection and pretreatment of data be in order to train neural network model, now,
On track, when reply board units are in each positioning region, corresponding wireless signal data, and the vehicle-mounted list of labelling are gathered
The region that unit is located.For example, table 3 is, when positioning region is two, a certain known location in classification 1 and classification 2 to be gathered respectively
Board units phase data table.After pretreatment is carried out to phase data, in being input to neural network model, and change each
Interneuronal weight, in further embodiments, can also change threshold values, constantly convert the position of board units or same
Position gathers again and repeat the above steps, until the neural network model meets the convergence of preset rules, so as to finally determine
Each interneuronal weight.In addition, in order that the result of training is not exclusively limited to training data, if with 100% number
According to training, the result for obtaining is a local minimum error, rather than global minima error, is instructed using 2/3 or so data
Practice, resulting result meets after convergence rule, training result is verified by other 1/3 data, acquired weight
Theoretically suitable for the global minima error of all data.It is used for verifying using 1/3 data, it can be ensured that in addition 2/3
Data train to be reliable.
Preferably, convergence rule has two：One be the number of times of training more than a certain value, such as 100,000,000 times；One is
Global error is less than some value.Carry out by one of them during training, or some reaches in two.
| Sequence number | Phase place 1 | Phase place 2 | Phase place 3 | Phase place 4 | Phase place 5 | Phase place 6 | Phase place 7 | Phase place 8 | Classification |
| 1 | 7665 | 566 | 1267 | 45678 | -876 | -9867 | -9876 | 876 | 1 |
| 2 | -9867 | -987 | 8765 | 19876 | 8873 | -123 | -222 | 134 | 2 |
Table 3
The parameters of neural network model in one embodiment are illustrated with reference to Fig. 4 and Fig. 5：Three array-element antenna arrays
Gather the phase data of wireless signal respectively, the phase data in pretreatment, phase data that array element A is gathered and array element B
The difference of the phase data for being gathered is p2, the difference of the phase data that the phase data that array element C is gathered and array element B are gathered
For p1.And, p1, p2 are normalized to into decimal, if x0=p1/m, x1=p2/m, m are normalized parameter.Then x0, x1 are made
For the input of artificial nerve network model.In addition, in the neural network model, the number of constructed input neuron is two
Individual, output neuron number is one, the middle number of plies of neuron is one layer, and the number of the intermediate layer neuron is six.
And, six intermediate layer neurons are input into interneuronal weight and are respectively relative to first：w01、w02、w03、w04、
W05, w06, six intermediate layer neurons are input into interneuronal weight and are respectively relative to second：w11、w12、w13、w14、
W15, w16, output neuron is respectively relative to the interneuronal weight in six intermediate layers：V1, v2, v3, v4, v5, v6, output
The output result of neuron is y.Finally the decision function of determination is：
Y=sum (vk*ok), k=1,2,3,4,5,6
Ok=sum (w0k*x0+w1k*x1)
Preferably, can be by the parameter configuration of the neural network model for training to FPGA or ARM.In addition, gathered data
Pretreatment also can carry out in FPGA or ARM.
Fig. 6 is the logic chart of roadside unit embodiment one of the present invention, and the roadside unit includes the acquisition module being sequentially connected
10th, pretreatment module 20 and Processing with Neural Network module 30, wherein, acquisition module 10 is used for collection from the wireless of electronic tag
Signal data, the data for being gathered include the amplitude and/or phase place and/or power of wireless signal；It is right that pretreatment module 20 is used for
The data for being gathered carry out pretreatment；Processing with Neural Network module 30 is used for pretreated data input to pre-building
In neural network model, and using the neural network model output as carrying electronic tag vehicle positioning result.
Preferably, the Processing with Neural Network module includes：Input module, for according to the number of the data to be gathered
It is determined that being input into the number of neuron, the number of the data to be gathered is related to the number of bay；Output module, for true
Determine the number of output neuron；Intermediate module, for determining the number of plies and intermediate layer neuron number in the middle of neuron；Adjustment mould
Block, for using the wireless signal data training god of the collection in training and the pretreated electronic tag from known location
Jing network modeies, and by constantly changing each interneuronal weight, until the neural network model meets preset rules
Convergence.
Preferably, acquisition module includes aerial array, radio-frequency transmitter, analogue signal processor and the modulus being sequentially connected
Transducer.Wherein, aerial array is used to receive the wireless signal from electronic tag, and the aerial array is preferably phased array day
Line；Radio-frequency transmitter is used to be demodulated the wireless signal that the aerial array is received；Processor is used for the radio frequency
Signal after receiver demodulation is processed；Analog-digital converter is used to carry out modulus turn to the signal after processor process
Change.
Preferably, pretreatment module 20 and Processing with Neural Network module 30 can be integrated in FPGA or ARM.
Fig. 7 is a kind of logic chart of the wireless location system embodiment one based on neutral net of the present invention, the embodiment
Wireless location system based on neutral net includes acquisition module and control module, wherein, acquisition module is used for collection from mesh
Target wireless signal data, the data for being gathered include the amplitude and/or phase place and/or power of wireless signal, the acquisition module
Including multiple antennas；Control module is used for the data to being gathered carries out pretreatment, and by pretreated data input in advance
In the neural network model first set up, and using the output of the neural network model as target positioning result.
Preferably, the control module includes：First input module, for being determined according to the number of the data to be gathered
The number of input neuron, the number of the data to be gathered is related to the number of bay；First output module, for true
Determine the number of output neuron；First intermediate module, for determining the number of plies and intermediate layer neuron number in the middle of neuron；First
Adjusting module, for training nerve net using the collection in training and the pretreated wireless signal data from known target
Network model, and by constantly changing each interneuronal weight, until the neural network model meets the convergence of preset rules.
The preferred embodiments of the present invention are the foregoing is only, the present invention is not limited to, for the skill of this area
For art personnel, the present invention can have various modifications and variations.It is all within the spirit and principles in the present invention, made any repair
Change, equivalent, improvement etc., should be included within scope of the presently claimed invention.
Claims (7)
1. a kind of wireless location method based on neutral net, it is characterised in that include：
S1. gather the wireless signal data from target, the data for being gathered include the amplitude and/or phase place of wireless signal and/
Or power；
S2. the data to being gathered carry out pretreatment；
S3. by pretreated data input to the neural network model for pre-building, and by the neural network model
Export the positioning result as target；
Wherein, neural network model is set up according to the following steps：
S31. the number of input neuron, the number and day of the data to be gathered are determined according to the number of the data to be gathered
The number of linear array unit is related；
S32. the number of output neuron is determined；
S33. the number of plies and intermediate layer neuron number in the middle of neuron is determined；
S34. the wireless signal data from known target are gathered in training, and the data to being gathered carry out pretreatment；
S35. the pretreated data of a part of step S34 institute are used for training neural network model, and it is each by constantly modification
Interneuronal weight and threshold values, until the neural network model meets the convergence of preset rules.
2. the wireless location method based on neutral net according to claim 1, it is characterised in that
It is single by the region division that need to be positioned into M predeterminable area when the output type of the neural network model is positioning region
Individual output neuron has two class output states, therefore the output of neural network model has 2NClass output state, wherein N are output
The number of neuron, the output state per Connectionist model is empty or corresponding with a predeterminable area, if output state
It is corresponding with a predeterminable area, that is, determine target in corresponding predeterminable area.
3. the wireless location method based on neutral net according to claim 1, it is characterised in that
When the output type of the neural network model is relative coordinates, the number of the output neuron is 2, the output god
Jing units are output as coordinate figure.
4. the wireless location method based on neutral net according to any one of claim 1-3, it is characterised in that the step
Rapid S2 includes：
Wireless signal to being gathered carries out pretreatment, obtains instant amplitude value；Or；
Wireless signal to being gathered in Preset Time carries out pretreatment, obtains the average power content of the wireless signal；Or；
Wireless signal to being gathered in Preset Time carries out pretreatment, obtains the phase place of wireless signal between different antenna element
Difference.
5. a kind of roadside unit, it is characterised in that include：
Acquisition module, for gathering the wireless signal data from electronic tag, the data for being gathered include the width of wireless signal
Degree and/or phase place and/or power；
Pretreatment module, for carrying out pretreatment to the data for being gathered；
Processing with Neural Network module, for by pretreated data input to the neural network model for pre-building, and will
Positioning result of the output of the neural network model as the vehicle for carrying electronic tag；
Wherein, the Processing with Neural Network module includes：
Input module, for determining the number of input neuron, the data to be gathered according to the number of the data to be gathered
Number it is related to the number of bay；
Output module, for determining the number of output neuron；
Intermediate module, for determining the number of plies and intermediate layer neuron number in the middle of neuron；
Adjusting module, for using the wireless communication of collection and the pretreated electronic tag from known location in training
Number trains neural network model, and by constantly changing each interneuronal weight and threshold values, until the neutral net
Model meets the convergence of preset rules.
6. roadside unit according to claim 5, it is characterised in that the acquisition module includes：
Aerial array, for receiving the wireless signal from electronic tag, the aerial array is phased array antenna；
Radio-frequency transmitter, for being demodulated to the wireless signal that the aerial array is received；
Analogue signal processor, for processing the signal after radio-frequency transmitter demodulation；
Analog-digital converter, for carrying out analog digital conversion to the signal after processor process.
7. a kind of wireless location system based on neutral net, it is characterised in that include：
Acquisition module, for collection from target wireless signal data, the amplitude of data for being gathered including wireless signal and/
Or phase place and/or power, the acquisition module is multiple antennas；
Control module, for the data for being gathered to be carried out with pretreatment, and by pretreated data input to pre-building
In neural network model, and using the output of the neural network model as target positioning result；
Wherein, the control module includes：
First input module, for determining the number of input neuron according to the number of the data to be gathered, to be gathered
The number of data is related to the number of bay；
First output module, for determining the number of output neuron；
First intermediate module, for determining the number of plies and intermediate layer neuron number in the middle of neuron；
First adjusting module, for being instructed using the collection in training and the pretreated wireless signal data from known target
Practice neural network model, and by constantly changing each interneuronal weight and threshold values, until the neural network model meets
The convergence of preset rules.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410447334.3A CN104569907B (en) | 2014-09-04 | 2014-09-04 | Wireless positioning method and system based on neural network and road side unit |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410447334.3A CN104569907B (en) | 2014-09-04 | 2014-09-04 | Wireless positioning method and system based on neural network and road side unit |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104569907A CN104569907A (en) | 2015-04-29 |
| CN104569907B true CN104569907B (en) | 2017-05-17 |
Family
ID=53086428
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410447334.3A Active CN104569907B (en) | 2014-09-04 | 2014-09-04 | Wireless positioning method and system based on neural network and road side unit |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104569907B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104951832B (en) * | 2015-06-05 | 2017-07-11 | 大连理工大学 | A kind of car networking roadside unit Optimization deployment method based on artificial fish-swarm algorithm |
| CN106355245B (en) * | 2016-09-12 | 2018-09-21 | 哈尔滨工业大学 | A kind of Pattern Synthesis of Antenna Array method based on neural network algorithm |
| WO2018204672A1 (en) | 2017-05-03 | 2018-11-08 | Oshea Timothy James | Learning radio signals using radio signal transformers |
| CN109696698B (en) * | 2019-03-05 | 2021-03-12 | 湖南国科微电子股份有限公司 | Navigation positioning prediction method and device, electronic equipment and storage medium |
| CN109922427B (en) * | 2019-03-06 | 2020-09-11 | 东南大学 | Intelligent wireless positioning system and method using large-scale array antenna |
| CN112288897B (en) * | 2020-09-30 | 2022-08-30 | 深圳市金溢科技股份有限公司 | Vehicle-mounted unit positioning method based on deep learning and road side unit |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1930490A (en) * | 2003-11-25 | 2007-03-14 | 麦卡利斯有限公司 | Object detection method and apparatus |
| CN101099090A (en) * | 2005-01-11 | 2008-01-02 | 极星公司 | Positioning device and method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8041333B2 (en) * | 2007-06-14 | 2011-10-18 | Broadcom Corporation | Method and system for 60 GHz antenna adaptation and user coordination based on base station beacons |
| US7800541B2 (en) * | 2008-03-31 | 2010-09-21 | Golba Llc | Methods and systems for determining the location of an electronic device |
-
2014
- 2014-09-04 CN CN201410447334.3A patent/CN104569907B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1930490A (en) * | 2003-11-25 | 2007-03-14 | 麦卡利斯有限公司 | Object detection method and apparatus |
| CN101099090A (en) * | 2005-01-11 | 2008-01-02 | 极星公司 | Positioning device and method |
Non-Patent Citations (1)
| Title |
|---|
| 《车联网环境下交通信息采集与处理方法研究》;周户星;《中国博士学位论文全文数据库 工程科技II辑》;20130815(第08期);正文第1-2、11、18-19、26、29、36、40、53-55、58、81-82、86-87页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104569907A (en) | 2015-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104569907B (en) | Wireless positioning method and system based on neural network and road side unit | |
| Kusiak et al. | Modeling and prediction of rainfall using radar reflectivity data: A data-mining approach | |
| CN110160541B (en) | Method and device for reconstructing motion trail, storage medium and electronic device | |
| CN102169584B (en) | Remote sensing image change detection method based on watershed and treelet algorithms | |
| CN105975913B (en) | Road network extraction method based on adaptive cluster learning | |
| CN106680775A (en) | Method and system for automatically identifying radar signal modulation modes | |
| Han et al. | Groundwater level prediction using a SOM-aided stepwise cluster inference model | |
| CN101448267A (en) | Wireless sensor network node coverage optimization method based on particle swarm algorithm | |
| CN101459915A (en) | Wireless sensor network node coverage optimization method based on genetic algorithm | |
| CN103885030A (en) | Locating method of mobile node in wireless sensor network | |
| CN106485924A (en) | The acquisition methods of dynamic traffic road network figure and device | |
| CN110490507B (en) | Method, device and equipment for detecting newly added line of logistics network | |
| CN108369682A (en) | Method and system for the stroke performance for evaluating driver | |
| CN105389817A (en) | Two-time phase remote sensing image change detection method | |
| CN108475359A (en) | Method and system for the stroke performance for evaluating driver | |
| CN102853836A (en) | Feedback weight fusion method based on track quality | |
| CN109543704A (en) | The multi-object clustering method, apparatus of automobile and trailer-mounted radar | |
| CN102162847A (en) | High-efficiency track correlation method based on singular value decomposition | |
| CN104535960A (en) | Indoor rapid positioning method based on RFID | |
| CN104469932A (en) | Position fingerprint positioning method based on support vector machine | |
| CN106803128A (en) | The Forecasting Methodology and device of photovoltaic generation | |
| CN104039008B (en) | A kind of hybrid locating method | |
| CN107437147A (en) | Reduce the vehicle travel risk dynamic assessment method and its system of freight logistics scene | |
| CN108369683A (en) | Method and system for the stroke performance for evaluating driver | |
| CN107490783A (en) | A kind of RFID localization methods based on ELM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |